refactor: 检查点加载重构，路径替代对象传递

- model: nn.Module -> model_fn 工厂函数，spawn 边界只传字符串
- Trainer.train(resume_dir=path) — Checkpoint 不再通过 pickle 传递
- TrainContextBuilder.with_resume_dir(path) — 自动检测 meta.json 分流 resume/from-scratch
- CheckpointCallback: 拆分 state_dict 收集（全 rank）与磁盘写入（rank-0），修复 FSDP 死锁
- serialization: load_torch 支持 broadcast，消除 _load_extra/_load_torch_broadcast
- optimizer/scheduler 恢复逻辑内联到 build()，在 executor.prepare() 之后执行
- pyproject.toml: ruff exclude build/ 避免 CI 扫描构建产物

tests/trainer/test_trainer.py

@@ -9,7 +9,7 @@ def test_different_batch_sizes(base_test_env, random_dataset, train_config_facto
 
     for batch_per_device in batch_sizes:
         train_config = train_config_factory(
-            model=base_test_env["model"],
+            model_fn=lambda: base_test_env["model"],
             dataset=random_dataset,
             test_dir=base_test_env["test_dir"],
             device=base_test_env["device"],
@@ -25,7 +25,7 @@ def test_gradient_accumulation(base_test_env, random_dataset, train_config_facto
 
     for grad_accum_steps in grad_accum_steps_list:
         train_config = train_config_factory(
-            model=base_test_env["model"],
+            model_fn=lambda: base_test_env["model"],
             dataset=random_dataset,
             test_dir=base_test_env["test_dir"],
             device=base_test_env["device"],
@@ -50,7 +50,7 @@ def test_memory_efficient_training(base_test_env, random_dataset, train_config_f
 
     for config in small_batch_configs:
         train_config = train_config_factory(
-            model=base_test_env["model"],
+            model_fn=lambda: base_test_env["model"],
             dataset=random_dataset,
             test_dir=base_test_env["test_dir"],
             device=base_test_env["device"],